Note Check the solidDB Release Notes for any limitations that are associated with using a grid in the current release.
Shutting down a grid is more complicated than shutting down a single node.
A functional grid with a write load has multiple asynchronous replication operations that run between grid nodes. If you attempt to shut down a grid by shutting down each of the individual nodes, the grid might attempt to restart nodes and to re-balance the workload of the grid onto nodes that are not yet shut down.
In an uncontrolled shut down or system power outage, nodes can be left in different states. Although a grid is designed to recover from outages, when you restart the grid, there is likely to be a delay before the grid can become active and serve traffic with full performance. Significant processing overhead and unnecessary CPU activity are likely as the grid attempts to restore the interrupted replication operations.
From an application perspective, the consequences of an uncontrolled shut down are likely to be slow startup, poor performance at startup, and potential loss of recently written data.
Shut down a grid safely, by completing the following tasks:
1 Turn off the workload in the applications that are communicating with the grid.
2 Check that there is no workload in the grid by running the following command on any grid node:
ADMIN COMMAND 'pmon'
3 Run the following command by using a grid connection:
ADMIN COMMAND 'grid shutdown'
The 'grid shutdown' command completes the following tasks:
▪ Passes the command execution to the node that is the grid leader. This task is only possible if the command is run from a grid connection.
▪ Informs the node controller that the grid is being shut down (to prevent nodes being restarted), see Node controller.
▪ Disables the Grid Availability Monitor (GAM) on all nodes (to prevent the corrective actions that are associated with grid node failure), see Grid Availability Manager.
▪ Shuts down all grid nodes. The grid leader is shut down last.
4 Find and shut down any remaining nodes.
Shutting down the grid shuts down only the nodes that are currently part of the grid. There might be grid nodes that were started but never joined to the grid that must be shut down separately. If grid orchestration is handled by an external program, you might need to consult the logs of that program.
As a result of network failures, it is possible to have some grid nodes that remain permanently disconnected from grid but are still active. You should find and shut down these nodes to avoid interference if these nodes become reachable.